Reproducibility: Fraud is not the big problem

Journal name:
Nature
Volume:
505,
Page:
483
Date published:
DOI:
doi:10.1038/505483b
Published online

Whistle-blowers risk a huge personal backlash in exposing scientific misconduct (see, for example, D. Soeken Nature 505, 26; 2014), but they can hope to correct only a tiny percentage of the published literature.

Since 1980, when MEDLINE started categorizing retractions, there have been 6,119 retracted papers, amounting to 0.03% of the 17.8 million published. Even if the majority of these retractions arise from misconduct (see, for example, F. C. Fang et al. Proc. Natl Acad. Sci. USA http://doi.org/jf5; 2012), this still affects only a very small proportion of the literature overall.

From alarming estimates derived from studies by Bayer (F. Prinz et al. Nature Rev. Drug Discov. 10, 712; 2011) and Amgen (C. G. Begley and L. M. Ellis Nature 483, 531533; 2012) that some 60–70% of biomedical research papers may contain irreproducible results, it would seem that our time would be better spent investigating experimental irreproducibility rather than hunting down fraudsters.

Author information

Affiliations

  1. Mendeley and the Reproducibility Initiative, California, USA.

    • William Gunn

Corresponding author

Correspondence to:

Author details

Additional data